FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
rtpenc.c
Go to the documentation of this file.
1 /*
2  * RTP output format
3  * Copyright (c) 2002 Fabrice Bellard
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * FFmpeg is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 #include "avformat.h"
23 #include "mpegts.h"
24 #include "internal.h"
25 #include "libavutil/mathematics.h"
26 #include "libavutil/random_seed.h"
27 #include "libavutil/opt.h"
28 
29 #include "rtpenc.h"
30 
31 static const AVOption options[] = {
33  { "payload_type", "Specify RTP payload type", offsetof(RTPMuxContext, payload_type), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 127, AV_OPT_FLAG_ENCODING_PARAM },
34  { "ssrc", "Stream identifier", offsetof(RTPMuxContext, ssrc), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, AV_OPT_FLAG_ENCODING_PARAM },
35  { "cname", "CNAME to include in RTCP SR packets", offsetof(RTPMuxContext, cname), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, AV_OPT_FLAG_ENCODING_PARAM },
36  { "seq", "Starting sequence number", offsetof(RTPMuxContext, seq), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 65535, AV_OPT_FLAG_ENCODING_PARAM },
37  { NULL },
38 };
39 
40 static const AVClass rtp_muxer_class = {
41  .class_name = "RTP muxer",
42  .item_name = av_default_item_name,
43  .option = options,
44  .version = LIBAVUTIL_VERSION_INT,
45 };
46 
47 #define RTCP_SR_SIZE 28
48 
49 static int is_supported(enum AVCodecID id)
50 {
51  switch(id) {
52  case AV_CODEC_ID_H261:
53  case AV_CODEC_ID_H263:
54  case AV_CODEC_ID_H263P:
55  case AV_CODEC_ID_H264:
56  case AV_CODEC_ID_HEVC:
59  case AV_CODEC_ID_MPEG4:
60  case AV_CODEC_ID_AAC:
61  case AV_CODEC_ID_MP2:
62  case AV_CODEC_ID_MP3:
65  case AV_CODEC_ID_PCM_S8:
70  case AV_CODEC_ID_PCM_U8:
72  case AV_CODEC_ID_AMR_NB:
73  case AV_CODEC_ID_AMR_WB:
74  case AV_CODEC_ID_VORBIS:
75  case AV_CODEC_ID_THEORA:
76  case AV_CODEC_ID_VP8:
79  case AV_CODEC_ID_ILBC:
80  case AV_CODEC_ID_MJPEG:
81  case AV_CODEC_ID_SPEEX:
82  case AV_CODEC_ID_OPUS:
83  return 1;
84  default:
85  return 0;
86  }
87 }
88 
90 {
91  RTPMuxContext *s = s1->priv_data;
92  int n;
93  AVStream *st;
94 
95  if (s1->nb_streams != 1) {
96  av_log(s1, AV_LOG_ERROR, "Only one stream supported in the RTP muxer\n");
97  return AVERROR(EINVAL);
98  }
99  st = s1->streams[0];
100  if (!is_supported(st->codec->codec_id)) {
101  av_log(s1, AV_LOG_ERROR, "Unsupported codec %s\n", avcodec_get_name(st->codec->codec_id));
102 
103  return -1;
104  }
105 
106  if (s->payload_type < 0) {
107  /* Re-validate non-dynamic payload types */
108  if (st->id < RTP_PT_PRIVATE)
109  st->id = ff_rtp_get_payload_type(s1, st->codec, -1);
110 
111  s->payload_type = st->id;
112  } else {
113  /* private option takes priority */
114  st->id = s->payload_type;
115  }
116 
118  s->timestamp = s->base_timestamp;
119  s->cur_timestamp = 0;
120  if (!s->ssrc)
121  s->ssrc = av_get_random_seed();
122  s->first_packet = 1;
125  /* Round the NTP time to whole milliseconds. */
126  s->first_rtcp_ntp_time = (s1->start_time_realtime / 1000) * 1000 +
128  // Pick a random sequence start number, but in the lower end of the
129  // available range, so that any wraparound doesn't happen immediately.
130  // (Immediate wraparound would be an issue for SRTP.)
131  if (s->seq < 0) {
132  if (s1->flags & AVFMT_FLAG_BITEXACT) {
133  s->seq = 0;
134  } else
135  s->seq = av_get_random_seed() & 0x0fff;
136  } else
137  s->seq &= 0xffff; // Use the given parameter, wrapped to the right interval
138 
139  if (s1->packet_size) {
140  if (s1->pb->max_packet_size)
141  s1->packet_size = FFMIN(s1->packet_size,
142  s1->pb->max_packet_size);
143  } else
144  s1->packet_size = s1->pb->max_packet_size;
145  if (s1->packet_size <= 12) {
146  av_log(s1, AV_LOG_ERROR, "Max packet size %d too low\n", s1->packet_size);
147  return AVERROR(EIO);
148  }
149  s->buf = av_malloc(s1->packet_size);
150  if (!s->buf) {
151  return AVERROR(ENOMEM);
152  }
153  s->max_payload_size = s1->packet_size - 12;
154 
155  s->max_frames_per_packet = 0;
156  if (s1->max_delay > 0) {
157  if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
159  if (!frame_size)
160  frame_size = st->codec->frame_size;
161  if (frame_size == 0) {
162  av_log(s1, AV_LOG_ERROR, "Cannot respect max delay: frame size = 0\n");
163  } else {
167  (AVRational){ frame_size, st->codec->sample_rate },
168  AV_ROUND_DOWN);
169  }
170  }
171  if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
172  /* FIXME: We should round down here... */
173  if (st->avg_frame_rate.num > 0 && st->avg_frame_rate.den > 0) {
175  (AVRational){1, 1000000},
176  av_inv_q(st->avg_frame_rate));
177  } else
178  s->max_frames_per_packet = 1;
179  }
180  }
181 
182  avpriv_set_pts_info(st, 32, 1, 90000);
183  switch(st->codec->codec_id) {
184  case AV_CODEC_ID_MP2:
185  case AV_CODEC_ID_MP3:
186  s->buf_ptr = s->buf + 4;
187  break;
190  break;
191  case AV_CODEC_ID_MPEG2TS:
192  n = s->max_payload_size / TS_PACKET_SIZE;
193  if (n < 1)
194  n = 1;
195  s->max_payload_size = n * TS_PACKET_SIZE;
196  s->buf_ptr = s->buf;
197  break;
198  case AV_CODEC_ID_H264:
199  /* check for H.264 MP4 syntax */
200  if (st->codec->extradata_size > 4 && st->codec->extradata[0] == 1) {
201  s->nal_length_size = (st->codec->extradata[4] & 0x03) + 1;
202  }
203  break;
204  case AV_CODEC_ID_HEVC:
205  /* Only check for the standardized hvcC version of extradata, keeping
206  * things simple and similar to the avcC/H264 case above, instead
207  * of trying to handle the pre-standardization versions (as in
208  * libavcodec/hevc.c). */
209  if (st->codec->extradata_size > 21 && st->codec->extradata[0] == 1) {
210  s->nal_length_size = (st->codec->extradata[21] & 0x03) + 1;
211  }
212  break;
213  case AV_CODEC_ID_VORBIS:
214  case AV_CODEC_ID_THEORA:
215  if (!s->max_frames_per_packet) s->max_frames_per_packet = 15;
216  s->max_frames_per_packet = av_clip(s->max_frames_per_packet, 1, 15);
217  s->max_payload_size -= 6; // ident+frag+tdt/vdt+pkt_num+pkt_length
218  s->num_frames = 0;
219  goto defaultcase;
221  /* Due to a historical error, the clock rate for G722 in RTP is
222  * 8000, even if the sample rate is 16000. See RFC 3551. */
223  avpriv_set_pts_info(st, 32, 1, 8000);
224  break;
225  case AV_CODEC_ID_OPUS:
226  if (st->codec->channels > 2) {
227  av_log(s1, AV_LOG_ERROR, "Multistream opus not supported in RTP\n");
228  goto fail;
229  }
230  /* The opus RTP RFC says that all opus streams should use 48000 Hz
231  * as clock rate, since all opus sample rates can be expressed in
232  * this clock rate, and sample rate changes on the fly are supported. */
233  avpriv_set_pts_info(st, 32, 1, 48000);
234  break;
235  case AV_CODEC_ID_ILBC:
236  if (st->codec->block_align != 38 && st->codec->block_align != 50) {
237  av_log(s1, AV_LOG_ERROR, "Incorrect iLBC block size specified\n");
238  goto fail;
239  }
240  if (!s->max_frames_per_packet)
241  s->max_frames_per_packet = 1;
242  s->max_frames_per_packet = FFMIN(s->max_frames_per_packet,
243  s->max_payload_size / st->codec->block_align);
244  goto defaultcase;
245  case AV_CODEC_ID_AMR_NB:
246  case AV_CODEC_ID_AMR_WB:
247  if (!s->max_frames_per_packet)
248  s->max_frames_per_packet = 12;
249  if (st->codec->codec_id == AV_CODEC_ID_AMR_NB)
250  n = 31;
251  else
252  n = 61;
253  /* max_header_toc_size + the largest AMR payload must fit */
254  if (1 + s->max_frames_per_packet + n > s->max_payload_size) {
255  av_log(s1, AV_LOG_ERROR, "RTP max payload size too small for AMR\n");
256  goto fail;
257  }
258  if (st->codec->channels != 1) {
259  av_log(s1, AV_LOG_ERROR, "Only mono is supported\n");
260  goto fail;
261  }
262  case AV_CODEC_ID_AAC:
263  s->num_frames = 0;
264  default:
265 defaultcase:
266  if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
267  avpriv_set_pts_info(st, 32, 1, st->codec->sample_rate);
268  }
269  s->buf_ptr = s->buf;
270  break;
271  }
272 
273  return 0;
274 
275 fail:
276  av_freep(&s->buf);
277  return AVERROR(EINVAL);
278 }
279 
280 /* send an rtcp sender report packet */
281 static void rtcp_send_sr(AVFormatContext *s1, int64_t ntp_time, int bye)
282 {
283  RTPMuxContext *s = s1->priv_data;
284  uint32_t rtp_ts;
285 
286  av_dlog(s1, "RTCP: %02x %"PRIx64" %x\n", s->payload_type, ntp_time, s->timestamp);
287 
288  s->last_rtcp_ntp_time = ntp_time;
289  rtp_ts = av_rescale_q(ntp_time - s->first_rtcp_ntp_time, (AVRational){1, 1000000},
290  s1->streams[0]->time_base) + s->base_timestamp;
291  avio_w8(s1->pb, RTP_VERSION << 6);
292  avio_w8(s1->pb, RTCP_SR);
293  avio_wb16(s1->pb, 6); /* length in words - 1 */
294  avio_wb32(s1->pb, s->ssrc);
295  avio_wb64(s1->pb, NTP_TO_RTP_FORMAT(ntp_time));
296  avio_wb32(s1->pb, rtp_ts);
297  avio_wb32(s1->pb, s->packet_count);
298  avio_wb32(s1->pb, s->octet_count);
299 
300  if (s->cname) {
301  int len = FFMIN(strlen(s->cname), 255);
302  avio_w8(s1->pb, (RTP_VERSION << 6) + 1);
303  avio_w8(s1->pb, RTCP_SDES);
304  avio_wb16(s1->pb, (7 + len + 3) / 4); /* length in words - 1 */
305 
306  avio_wb32(s1->pb, s->ssrc);
307  avio_w8(s1->pb, 0x01); /* CNAME */
308  avio_w8(s1->pb, len);
309  avio_write(s1->pb, s->cname, len);
310  avio_w8(s1->pb, 0); /* END */
311  for (len = (7 + len) % 4; len % 4; len++)
312  avio_w8(s1->pb, 0);
313  }
314 
315  if (bye) {
316  avio_w8(s1->pb, (RTP_VERSION << 6) | 1);
317  avio_w8(s1->pb, RTCP_BYE);
318  avio_wb16(s1->pb, 1); /* length in words - 1 */
319  avio_wb32(s1->pb, s->ssrc);
320  }
321 
322  avio_flush(s1->pb);
323 }
324 
325 /* send an rtp packet. sequence number is incremented, but the caller
326  must update the timestamp itself */
327 void ff_rtp_send_data(AVFormatContext *s1, const uint8_t *buf1, int len, int m)
328 {
329  RTPMuxContext *s = s1->priv_data;
330 
331  av_dlog(s1, "rtp_send_data size=%d\n", len);
332 
333  /* build the RTP header */
334  avio_w8(s1->pb, RTP_VERSION << 6);
335  avio_w8(s1->pb, (s->payload_type & 0x7f) | ((m & 0x01) << 7));
336  avio_wb16(s1->pb, s->seq);
337  avio_wb32(s1->pb, s->timestamp);
338  avio_wb32(s1->pb, s->ssrc);
339 
340  avio_write(s1->pb, buf1, len);
341  avio_flush(s1->pb);
342 
343  s->seq = (s->seq + 1) & 0xffff;
344  s->octet_count += len;
345  s->packet_count++;
346 }
347 
348 /* send an integer number of samples and compute time stamp and fill
349  the rtp send buffer before sending. */
351  const uint8_t *buf1, int size, int sample_size_bits)
352 {
353  RTPMuxContext *s = s1->priv_data;
354  int len, max_packet_size, n;
355  /* Calculate the number of bytes to get samples aligned on a byte border */
356  int aligned_samples_size = sample_size_bits/av_gcd(sample_size_bits, 8);
357 
358  max_packet_size = (s->max_payload_size / aligned_samples_size) * aligned_samples_size;
359  /* Not needed, but who knows. Don't check if samples aren't an even number of bytes. */
360  if ((sample_size_bits % 8) == 0 && ((8 * size) % sample_size_bits) != 0)
361  return AVERROR(EINVAL);
362  n = 0;
363  while (size > 0) {
364  s->buf_ptr = s->buf;
365  len = FFMIN(max_packet_size, size);
366 
367  /* copy data */
368  memcpy(s->buf_ptr, buf1, len);
369  s->buf_ptr += len;
370  buf1 += len;
371  size -= len;
372  s->timestamp = s->cur_timestamp + n * 8 / sample_size_bits;
373  ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 0);
374  n += (s->buf_ptr - s->buf);
375  }
376  return 0;
377 }
378 
380  const uint8_t *buf1, int size)
381 {
382  RTPMuxContext *s = s1->priv_data;
383  int len, count, max_packet_size;
384 
385  max_packet_size = s->max_payload_size;
386 
387  /* test if we must flush because not enough space */
388  len = (s->buf_ptr - s->buf);
389  if ((len + size) > max_packet_size) {
390  if (len > 4) {
391  ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 0);
392  s->buf_ptr = s->buf + 4;
393  }
394  }
395  if (s->buf_ptr == s->buf + 4) {
396  s->timestamp = s->cur_timestamp;
397  }
398 
399  /* add the packet */
400  if (size > max_packet_size) {
401  /* big packet: fragment */
402  count = 0;
403  while (size > 0) {
404  len = max_packet_size - 4;
405  if (len > size)
406  len = size;
407  /* build fragmented packet */
408  s->buf[0] = 0;
409  s->buf[1] = 0;
410  s->buf[2] = count >> 8;
411  s->buf[3] = count;
412  memcpy(s->buf + 4, buf1, len);
413  ff_rtp_send_data(s1, s->buf, len + 4, 0);
414  size -= len;
415  buf1 += len;
416  count += len;
417  }
418  } else {
419  if (s->buf_ptr == s->buf + 4) {
420  /* no fragmentation possible */
421  s->buf[0] = 0;
422  s->buf[1] = 0;
423  s->buf[2] = 0;
424  s->buf[3] = 0;
425  }
426  memcpy(s->buf_ptr, buf1, size);
427  s->buf_ptr += size;
428  }
429 }
430 
432  const uint8_t *buf1, int size)
433 {
434  RTPMuxContext *s = s1->priv_data;
435  int len, max_packet_size;
436 
437  max_packet_size = s->max_payload_size;
438 
439  while (size > 0) {
440  len = max_packet_size;
441  if (len > size)
442  len = size;
443 
444  s->timestamp = s->cur_timestamp;
445  ff_rtp_send_data(s1, buf1, len, (len == size));
446 
447  buf1 += len;
448  size -= len;
449  }
450 }
451 
452 /* NOTE: size is assumed to be an integer multiple of TS_PACKET_SIZE */
454  const uint8_t *buf1, int size)
455 {
456  RTPMuxContext *s = s1->priv_data;
457  int len, out_len;
458 
459  while (size >= TS_PACKET_SIZE) {
460  len = s->max_payload_size - (s->buf_ptr - s->buf);
461  if (len > size)
462  len = size;
463  memcpy(s->buf_ptr, buf1, len);
464  buf1 += len;
465  size -= len;
466  s->buf_ptr += len;
467 
468  out_len = s->buf_ptr - s->buf;
469  if (out_len >= s->max_payload_size) {
470  ff_rtp_send_data(s1, s->buf, out_len, 0);
471  s->buf_ptr = s->buf;
472  }
473  }
474 }
475 
476 static int rtp_send_ilbc(AVFormatContext *s1, const uint8_t *buf, int size)
477 {
478  RTPMuxContext *s = s1->priv_data;
479  AVStream *st = s1->streams[0];
480  int frame_duration = av_get_audio_frame_duration(st->codec, 0);
481  int frame_size = st->codec->block_align;
482  int frames = size / frame_size;
483 
484  while (frames > 0) {
485  int n = FFMIN(s->max_frames_per_packet - s->num_frames, frames);
486 
487  if (!s->num_frames) {
488  s->buf_ptr = s->buf;
489  s->timestamp = s->cur_timestamp;
490  }
491  memcpy(s->buf_ptr, buf, n * frame_size);
492  frames -= n;
493  s->num_frames += n;
494  s->buf_ptr += n * frame_size;
495  buf += n * frame_size;
496  s->cur_timestamp += n * frame_duration;
497 
498  if (s->num_frames == s->max_frames_per_packet) {
499  ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 1);
500  s->num_frames = 0;
501  }
502  }
503  return 0;
504 }
505 
507 {
508  RTPMuxContext *s = s1->priv_data;
509  AVStream *st = s1->streams[0];
510  int rtcp_bytes;
511  int size= pkt->size;
512 
513  av_dlog(s1, "%d: write len=%d\n", pkt->stream_index, size);
514 
515  rtcp_bytes = ((s->octet_count - s->last_octet_count) * RTCP_TX_RATIO_NUM) /
517  if ((s->first_packet || ((rtcp_bytes >= RTCP_SR_SIZE) &&
518  (ff_ntp_time() - s->last_rtcp_ntp_time > 5000000))) &&
519  !(s->flags & FF_RTP_FLAG_SKIP_RTCP)) {
520  rtcp_send_sr(s1, ff_ntp_time(), 0);
522  s->first_packet = 0;
523  }
524  s->cur_timestamp = s->base_timestamp + pkt->pts;
525 
526  switch(st->codec->codec_id) {
529  case AV_CODEC_ID_PCM_U8:
530  case AV_CODEC_ID_PCM_S8:
531  return rtp_send_samples(s1, pkt->data, size, 8 * st->codec->channels);
536  return rtp_send_samples(s1, pkt->data, size, 16 * st->codec->channels);
538  /* The actual sample size is half a byte per sample, but since the
539  * stream clock rate is 8000 Hz while the sample rate is 16000 Hz,
540  * the correct parameter for send_samples_bits is 8 bits per stream
541  * clock. */
542  return rtp_send_samples(s1, pkt->data, size, 8 * st->codec->channels);
544  return rtp_send_samples(s1, pkt->data, size,
546  case AV_CODEC_ID_MP2:
547  case AV_CODEC_ID_MP3:
548  rtp_send_mpegaudio(s1, pkt->data, size);
549  break;
552  ff_rtp_send_mpegvideo(s1, pkt->data, size);
553  break;
554  case AV_CODEC_ID_AAC:
555  if (s->flags & FF_RTP_FLAG_MP4A_LATM)
556  ff_rtp_send_latm(s1, pkt->data, size);
557  else
558  ff_rtp_send_aac(s1, pkt->data, size);
559  break;
560  case AV_CODEC_ID_AMR_NB:
561  case AV_CODEC_ID_AMR_WB:
562  ff_rtp_send_amr(s1, pkt->data, size);
563  break;
564  case AV_CODEC_ID_MPEG2TS:
565  rtp_send_mpegts_raw(s1, pkt->data, size);
566  break;
567  case AV_CODEC_ID_H264:
568  ff_rtp_send_h264(s1, pkt->data, size);
569  break;
570  case AV_CODEC_ID_H261:
571  ff_rtp_send_h261(s1, pkt->data, size);
572  break;
573  case AV_CODEC_ID_H263:
574  if (s->flags & FF_RTP_FLAG_RFC2190) {
575  int mb_info_size = 0;
576  const uint8_t *mb_info =
578  &mb_info_size);
579  if (!mb_info) {
580  av_log(s1, AV_LOG_ERROR, "failed to allocate side data\n");
581  return AVERROR(ENOMEM);
582  }
583  ff_rtp_send_h263_rfc2190(s1, pkt->data, size, mb_info, mb_info_size);
584  break;
585  }
586  /* Fallthrough */
587  case AV_CODEC_ID_H263P:
588  ff_rtp_send_h263(s1, pkt->data, size);
589  break;
590  case AV_CODEC_ID_HEVC:
591  ff_rtp_send_hevc(s1, pkt->data, size);
592  break;
593  case AV_CODEC_ID_VORBIS:
594  case AV_CODEC_ID_THEORA:
595  ff_rtp_send_xiph(s1, pkt->data, size);
596  break;
597  case AV_CODEC_ID_VP8:
598  ff_rtp_send_vp8(s1, pkt->data, size);
599  break;
600  case AV_CODEC_ID_ILBC:
601  rtp_send_ilbc(s1, pkt->data, size);
602  break;
603  case AV_CODEC_ID_MJPEG:
604  ff_rtp_send_jpeg(s1, pkt->data, size);
605  break;
606  case AV_CODEC_ID_OPUS:
607  if (size > s->max_payload_size) {
608  av_log(s1, AV_LOG_ERROR,
609  "Packet size %d too large for max RTP payload size %d\n",
610  size, s->max_payload_size);
611  return AVERROR(EINVAL);
612  }
613  /* Intentional fallthrough */
614  default:
615  /* better than nothing : send the codec raw data */
616  rtp_send_raw(s1, pkt->data, size);
617  break;
618  }
619  return 0;
620 }
621 
623 {
624  RTPMuxContext *s = s1->priv_data;
625 
626  /* If the caller closes and recreates ->pb, this might actually
627  * be NULL here even if it was successfully allocated at the start. */
628  if (s1->pb && (s->flags & FF_RTP_FLAG_SEND_BYE))
629  rtcp_send_sr(s1, ff_ntp_time(), 1);
630  av_freep(&s->buf);
631 
632  return 0;
633 }
634 
636  .name = "rtp",
637  .long_name = NULL_IF_CONFIG_SMALL("RTP output"),
638  .priv_data_size = sizeof(RTPMuxContext),
639  .audio_codec = AV_CODEC_ID_PCM_MULAW,
640  .video_codec = AV_CODEC_ID_MPEG4,
644  .priv_class = &rtp_muxer_class,
645 };